EP1033036B1 - Adaptive entropy coding in adaptive quantization framework for video signal coding systems and processes - Google Patents

Adaptive entropy coding in adaptive quantization framework for video signal coding systems and processes Download PDF

Info

Publication number
EP1033036B1
EP1033036B1 EP19980960467 EP98960467A EP1033036B1 EP 1033036 B1 EP1033036 B1 EP 1033036B1 EP 19980960467 EP19980960467 EP 19980960467 EP 98960467 A EP98960467 A EP 98960467A EP 1033036 B1 EP1033036 B1 EP 1033036B1
Authority
EP
European Patent Office
Prior art keywords
esc
quantizer
esc esc
value
run
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Not-in-force
Application number
EP19980960467
Other languages
German (de)
French (fr)
Other versions
EP1033036A1 (en
Inventor
Anurag Bist
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Lakestar Semi Inc
Original Assignee
Lakestar Semi Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority to US08/980,654 priority Critical patent/US6111914A/en
Priority to US980654 priority
Priority to US09/198,742 priority patent/US6118822A/en
Application filed by Lakestar Semi Inc filed Critical Lakestar Semi Inc
Priority to PCT/US1998/025155 priority patent/WO1999029115A1/en
Publication of EP1033036A1 publication Critical patent/EP1033036A1/en
Priority to US198742 priority
Application granted granted Critical
Publication of EP1033036B1 publication Critical patent/EP1033036B1/en
Anticipated expiration legal-status Critical
Not-in-force legal-status Critical Current

Links

Images

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/124Quantisation
    • H04N19/126Details of normalisation or weighting functions, e.g. normalisation matrices or variable uniform quantisers
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/60Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/124Quantisation
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/13Adaptive entropy coding, e.g. adaptive variable length coding [AVLC] or context adaptive binary arithmetic coding [CABAC]
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/90Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using coding techniques not provided for in groups H04N19/10-H04N19/85, e.g. fractals
    • H04N19/91Entropy coding, e.g. variable length coding [VLC] or arithmetic coding

Description

    BACKGROUND OF THE INVENTION 1. Field of the Invention
  • The present invention relates to devices and methods for coding video signals and more specifically to the coding using adaptive quantization with multiple quantizers and variable length code (VLC) tables. These devices and data compression methods can be readily adapted for H.263+ and MPEG-4 video standards.
  • 2. Description of Related Art
  • Video signals generally include data corresponding to one or more video frames, where each video frame is composed of an array of picture elements (pels). A typical color video frame can be of a variety of resolutions, one of which is the quarter common interface format (QCIF) resolution represented in Fig. 1, where a video frame is composed of over twenty-five thousand picture elements, arranged in an array of 144 pels x 176 pels, and divided in 8x8 pel blocks. Since each pel has to be characterized with a color (or hue) and luminance characteristics, these data may be represented with groups of four luminance pel blocks or two chrominance pel blocks called macroblocks. Thus, digital signals representing a sequence of video frame data, usually containing many video frames, have a large number of bits. However, the available storage space and bandwidth for transmitting such signals is limited. Therefore, compression (coding) processes are used to more efficiently transmit or store video data.
  • Compression of digital video signals for transmission or for storage has become widely practiced in a variety of contexts, especially in multimedia environments for video conferencing, video games, Internet image transmissions, digital TV. and the like. Coding and decoding is accomplished with coding processors which may be general computers, special hardware or multimedia boards and other suitable processing devices. Standards for compression processes have been developed by the International Telecommunication Union (ITU), which has developed H series standards used for real-time communications such as in videophones, and the International Organization for Standardization (ISO) which has developed the Motion Picture Experts Group (MPEG) series standards, such as MPEG-1, MPEG-2, MPEG-4 and MPEG-7.
  • Coding operations for video signals are typically performed with smaller samples of the video frame, such as the 16 x 16 pel sample composed of four 8 x 8 pel blocks shown in broken lines in Fig. 1. Compression processes typically involve quantization, in which sampled video signal data values, like color and luminance, are represented by, or are mapped onto, a fixed number of predefined quantizer values. The quantized signal is composed of quantizer values that are, in fact, approximations of the sampled video signal values. Therefore, the encoding of the video signal data onto a limited number of quantizer values necessarily produces some loss in accuracy and distortion of the signal after decoding process.
  • Methods to improve video signal transmission have been variously introduced. For example, WO-A-9634496 discloses a device for coding a video signal using a quantizer that is adapted to quantize the coefficient of a picture transformer to a non-zero quantization value (or level) that is lower than the most proximate quantization value, if the non-zero value is more efficient in terms of rate-distortion. EP-A-0 575 113 discloses a bit rate reduction apparatus which, in one embodiment, modifies data values by a selected offset value to enhance transmission. But neither WO-A96 34496 nor EP-A-0 575 113 discloses a method for selecting a quantizer set of quantization values for each portion of a video frame from a group of quantizer sets based on the distortion and bit rate characteristics of each quantizer set in the group.
  • EP-A-0 859 520, which falls under Article 54(3) EPC, describes a compression system and process that employs a group of quantizers (or sets of predefined quantizer values), and the selection of one of the quantizers for each video frame. However, EP-A-0 859 520 does not disclose how the quantizer values are to be stored.
  • Improvements in accuracy and distortion may be made by designing the quantizer set to include those quantizer values that have been found by experimentation to have the highest probability of being selected. However, the quantizer values that have the highest selection probability in one portion of a video frame are typically different from the quantizer values that have the highest selection probability in another portion of the frame, and, in a moving video context, such values differ from frame to frame as well,
  • A compression process of an input signal I, according to the H.263 and MPEG-4 standards, is generally represented by Fig. 2 for Intra blocks and by Fig. 3 for Inter blocks, and involves a discrete cosine transform (DCT) step 12, a quantization step 16, and a variable length coding step 18. Compressed and coded input video data signal is then transmitted on a fixed bandwidth communication link to a receiver (not shown). For Inter blocks, the process shown in Fig. 2 includes an additional motion compensation step 11, carried out prior to the DCT step 12. The motion compensation step 11 involves the summing operation in which a displaced block from a previous frame using motion vector from motion estimation is subtracted from the input video data signal I for a block, to provide a motion compensated difference block as the input to the DCT step 12.
  • In Figs. 2 and 3, the DCT step 12 is performed on the signal components associated with the luminance and color values for each processing sample block and provides an output for each block, representing components of each pel of a block as values within a defined range. The range of an output block may be, for example, (-2048, +2048), with the higher values typically positioned toward the top left corner of each block and the values decreasing in the direction of arrow 14 toward the lower right corner of each block.
  • In the quantization step 16, each DCT value in the block is mapped to a value within a fixed set of quantizer values, based on a quantization scheme. As discussed in more detail below, a preferred quantization scheme in accordance with H.263+ and MPEG-4 video standards involves a step of calculating new values for luminance and color components of each pel of a block. Following the calculation step, a selecting step is carried out, in which approximate values are selected from predefined quantizer values closest to the original input signal values.
  • In most low bit rate video coding schemes, a calculation step for calculating new values Y' in the quantization process involves a number of calculations to determine a level for Y data value of each pel in the block, using the following equation: level = INT[(Y + offset)/2QP] where INT is the integer value of the equation.
  • Y could be a value of luminance or color, offset is preferably -QP/2, and QP is a quantization coefficient representing quantization step size, preferably selected from the range 1 ≤ QP ≤ 31. The QP value is determined using well known linear bit rate control techniques and it depends on the number of bits allowed by the bit constraints of the system and on the desired image quality.
  • From the calculated level for each pel, a Y' value is determined, using the following equation: Y' = (2QP x level) + QP In this manner, a Y' value is calculated for each Y value in the block and each Y' value is a multiple of QP, based on the corresponding Y value.
  • In typical quantization systems, in accordance with the H.263 and MPEG-4 standards, the Y' value is used for selecting a quantizer value from a set of quantizer values represented by the standard quantizer R0, where each quantizer value is a multiple of QP. The standard quantizer R0 is defined as follows: R0 ∈ {0, ±3QP, ±5QP, ±7QP,...} Each Y' value is mapped onto the closest quantizer value selected from the set Ro and, thus, each block of Y values is mapped to a corresponding block of quantizer values.
  • For example, if the Y value for a particular pel is +110 and the selected QP value is 20, then the above equations would render the following values: offset = -QP/2 = -10 level = INT[(Y+offset)/2QP] = INT [2.5] = 2 Y' = (2QP) x level + QP = 40 x 2 + 20 = 100 = 5QP If the value of Y is chosen to be within the range -2048 ≤ Y ≤ 2048, then the quantizer R0 of estimated Y' values must have a QP-multiple which represents the lowest possible value (-2048) as well as the highest possible value (2048). Using the standard QP-multiples set R0, the quantizer R0 would have to extend to ± 101 QP and is represented as follows: R0 ∈ {0, ±3QP, ±5QP, ±7QP,... ±101QP,... ±255QP }
  • The calculated Y' value of 5QP happens to be identical to a QP-multiple value within the set: {0, ±3QP, ±5QP, ±7QP,... ±101QP,... ±255QP}. In this example the values above ±110 are not used and the Y value 110 of the pel was mapped to the Y' value 5QP. However, the accuracy of the coded information suffers to some extent, as illustrated by the fact that the Y' value of 100 differs from the original Y value of 110 by almost 10% of the Y value.
  • In the second example, with the same Y value of 110, as above, but with the QP value of 2: level = [(110-2/2)/2(2)] = [27.25] = 27 and Y'= 2(2) x 27 + 2 = 110 = 55QP. Thus, in the second example, the Y value 110 of the pel was mapped to the Y' value of 55QP and the calculated Y' value of 110 is equal to the original Y value.
  • With the relatively low QP value of 2 the number of possible QP-multiples is significantly greater than in the first example. In particular, with -2048 ≤ Y ≤ +2048 and with QP = 2, the standard quantizer R0 of possible Y' values uses the values up to +255QP, as follows: {0, ±3QP, ±5QP, ±7QP, ... ±255QP}. Thus, with the QP value of 2, the number of possible Y' values is increased with respect to the first example, in which the QP value was 20. Accordingly, the accuracy of the Y' value is greater in the second example than in the first example.
  • The number of bits required to code the possible Y' value is a function of the absolute value of the input signal I, and the higher the magnitude of the signal, the more bits it would take to represent the possible Y' values. Because 27 is significantly greater than 2, the number of bits needed to code the possible Y' value is greater in the second than in the first example. Therefore, the QP value should be selected based on the number of bits that can be transmitted over the transmission channel and the desired image quality, because more accurate coding (mapping) of pel values occurs with lower QP values. Thus, lower QP values may be used with certain portions of the video frame in which higher accuracy is desired, such as portions of the frame in which movement occurs, than in the other portions of the frame. Also, different QP values could be used for luminance and chrominance blocks.
  • As a result of the DCT step 12, the Y' values toward the lower right comer of the block will tend to be small and, if QP is high enough, the Y' values for many of the pels toward that corner will be zero, as it is shown in Fig. 4, where an "x" represents a value other than zero and "0" represents a value of zero.
  • After the quantization step 16, the pel values in the block are coded using a variable length coding 18 step with a suitable entropy coding scheme, such as Huffman coding. Huffman coding is preferred because of its low cost, high accuracy and efficiency. With this coding the pel values can be coded with a relatively short binary code sequence and the strings of zeroes (called Runs), which tend to occur toward the lower right comer of the quantized block, are coded together with the code for the next pel in one binary string. The efficiency can be further enhanced by coding the data for each pel in the zig-zag sequence, represented by the arrow 20 in Fig. 4.
  • At the receiver end, the Huffman-coded transmission is decoded to obtain the reconstructed values for the coded block of Y' values. Afterwards, calculations reversing the above-discussed calculations are performed on the Y' values to obtain representations of the original Y values. A VLC table used for the encoding and decoding in the H.263+ standard is given in Figures 5a and 5b.
  • OBJECTS AND SUMMARY OF THE INVENTION
  • It is an object of preferred embodiments of the present invention to provide an improved compression system and method which are capable of providing improved accuracy with less distortion and/or improved bit rate characteristics.
  • It is a further object of preferred embodiments of the invention to provide such a system and method which adapt to the statistical changes of the imagery in a video signal while maintaining the basic structure of the quantization processes of H 263+ and MPEG-4 standards.
  • These and other objects are achieved by a method according to the independent method claim and by a system according to the independent system claim. The invention creates a system and process for coding video signals employing an adaptive quantization scheme for quantizing the sampled data values by correlating each sampled data value to a value selected from a predefined set Rx of quantizer values. The quantizer values are multiples of a quantization coefficient QP, and the quantizer set is chosen from a group Ro - Rn of quantizer sets based on a distortion and bit rate characteristics of each quantizer value in the set Rx. At least one VLC table, having an entry for each pel event with a high probability of occurrence, is selected and the quantizer values are coded with a variable length adaptive entropy coding method using the VLC table. The VLC table is obtained in a training procedure using a training signal; based on a probability of occurrence of each event from the total number of occurrences of the event in the training signal. The training procedure may use several VLC tables, one for each quantizer Rx, for improved efficiency.
  • BRIEF DESCRIPTION OF THE DRAWINGS
  • The objects and features of the present invention, which are believed to be novel, are set forth with particularity in the appended claims. The present invention, both as to its organization and manner of operation, together with further objects and advantages, may best be understood by reference to the following description, taken in connection with the accompanying drawings.
  • Figure 1 is a representation of a typical video frame;
  • Figure 2 is a block diagram representing a compression process for coding video signal data for Intra blocks;
  • Figure 3 is a block diagram representing a compression process for coding video signal data for Inter blocks;
  • Figure 4 is a representation of an 8 x 8 pel block of a video frame, after the DCT and quantization steps of the process shown in Fig. 2;
  • Figures 5a and 5b are representations of the VLC table used in H.263+ standard;
  • Figure 6 is a representation of a table showing lengths of binary codes for non-last events, in an optimized 3-D VLC table obtained using the adaptive entropy coding of the present invention;
  • Figure 7 is a representation of a table showing the lengths of binary codes for last events, in an optimized 3-D VLC table obtained using the adaptive entropy coding of the present invention;
  • Figure 8 is a representation of a table showing pel values of an 8x8 input block;
  • Figure 9 is a representation of a table showing a quantized block which corresponds to the input block of Figure 8.
  • Figure 10 is a representation of a table showing a block which corresponds to the quantized and scanned input block of Figure 8.
  • Figure 11 is an operational block diagram of another preferred method of the present invention for performing the quantization step using an offset γ.
  • Figure 12 is a representation of a trellis of the present invention of decision levels which correspond to all possible choices of level values and run values of consecutive zeros.
  • Figure 13 is an operational block diagram of yet another preferred embodiment of the present invention for performing the quantization step searching through the trellis of Figure 12.
  • DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
  • The following detailed description is of the best presently contemplated mode of carrying out the invention. This description is not to be taken in a limiting sense, but is made merely for the purpose of illustrating general principles of embodiments of the invention. The scope of the invention is best defined by the appended claims.
  • Conventional coding systems, employing current H.263+ and MPEG-4 video standards, do not have the ability to change quantizers values to adapt to the changes in the input signal because there is only one quantizer R0.
  • In a preferred embodiment of the present invention a method is proposed where the quantization for each frame or, more preferably, for each frame portion, is carried out in the quantization step 16 by selecting the most appropriate quantizer from a group of predefined quantizers R0 - Rn. In this manner, the system or process is capable of changing to a different quantizer for different video frame or, preferably, for different portion of the same frame to achieve higher degree of accuracy and a lower bit rate. This method is partially described in the copending U.S. Patent Application by the same inventor entitled "Video Signal Coding Systems and Processes Using Adaptive Quantization," Serial No. 08/799,995, filed on February 13, 1997.
  • The quantizer selection is based on a determination of which quantizer provides the best distortion and bit rate characteristics for the portion of the video signal being coded. The selection is based on a formula which takes into account both the distortion and bit rate. A similar formula, based on both distortion and bit rate characteristics, is used to select the particular quantizer value within the quantizer for each video signal value being coded.
  • Multiple quantizers R0 - Rn, each having a set of quantizer values that are most likely to closely estimate the actual video signal values, are derived by experimentation. Different sets may be derived for different QP values.
  • For example, the U.S. Patent Application with serial number 08/799,995 proposes the following group of quantizers R0- Rn for quantization step QP: R0 = {0, ±3QP, ±5QP, ±7QP,...±255QP} R1 = {0, ±3QP, ±6QP, ±7QP,...±255QP} R2 = {0, ±3QP, ±4QP, ±7QP,... ±255QP} R3 = {0, ±2QP, ±6QP, ±7QP,... ±255QP} R4 = {0, ±2QP, ±5QP, ±7QP,... ±255QP} R5 = {0, ±2QP, ±4QP, ±7QP,... ±255QP} R6 = {0, ±2QP, ±3QP, ±7QP,... ±255QP} R7 = {0, ±4QP, ±6QP, ±7QP,... ±255QP} R8 = {0, ±4QP, ±5QP, ±7QP,... ±255QP} R9 = {0, ±1QP, ±6QP, ±7QP,... ±255QP} R10 = {0, ±1QP, ±5QP, ±7QP,... ±255QP} R11 = {0, ±1QP, ±4QP, ±7QP,... ±255QP} R12 = {0, ±1QP, ±3QP, ±7QP,... ±255QP} R13 = {0, ±1QP, ±2QP, ±7QP,... ±255QP} R14 = {0, ±5QP, ±6QP, ±7QP,... ±255QP}
  • The lowest two non-zero quantizer values are different in all 15 quantizers. By having multiple quantizer values for the lowest two non-zero level indexes of each quantizer a very efficient quantization is performed in the region where most of the input levels lie. In the adaptive quantization scheme of the present invention a different group of quantizers is defined by their quantizer values. The following group of quantizers R0 - Rn is preferred in the present invention because it produces more reliable results and is more efficient: R0 = {0, ±3QP-α, ±5QP-α, ±7QP-α ... ±255QP-α} R1 = {0, ±3QP-α, ±6QP-α, ±7QP-α ... ±255QP-α} R2 = {0, ±3QP-α, ±4QP-α, ±7QP-α, ... ±255QP-α} R3 = {0, ±2QP-α, ±6QP-α, ±7QP-α, ... ±255QP-α} R4 = {0, ±2QP-α, ±5QP-α, ±7QP-α, ... ±255QP-α} R5 = {0, ±2QP-α, ±4QP-α, ±7QP-α, ... ±255QP-α} R6 = {0, ±2QP-α, ±3QP-α ±7QP-α, ... ±255QP-α} R7 = {0, ±4QP-α, ±6QP-α, ±7QP-α,...±255QP-α} R8 = {0, ±4QP-α, ±5QP-α, ±7QP-α,...±255QP-α} R9 = {0, ±1QP-α, ±6QP-α, ±7QP-α, ... ±255QP-α} R10 = {0, ±1QP-α, ±5QP-α, ±7QP-α...±255QP-α} R11 = {0, ±1QP-α, ±4QP-α ±7QP-α, ... ±255QP-α} R12 = {0, ±1QP-α, ±3QP-α, ±7QP-α,...±255QP-α} R13 = {0, ±1QP-α, ±2QP-α, ±7QP-α,...±255QP-α} R14 = {0, ±5QP-α, ±6QP-,α ±7QP-α,...±255QP-α}    where the quantization step size QP can be any number between 1 and 31. α = 1 if QP is an even number and α = 0 if QP is an odd number. The quantizer values of each quantizer are saved in an appropriate element of a one-dimensional Quantizer Value Array which corresponds to that quantizer, and can be retrieved using a quantizer value Level index.
  • While there may be any number of quantizers R0 - Rn, which differ from each other to any suitable degree, it has been found that significantly improved accuracy and bit rate can be achieved with quantizers that differ from each other at the lower multiples of QP. Accordingly, the above sets R0 - R14 differ from each other at the two lowest non-zero multiples of QP, i.e., below 7QP and above -7QP, and the quantizer sets R0 - R14 represent all possible sequences of multiples of QP between -7QP and 7QP. For multiples of QP greater than 6 and less than -6 the above sets are identical. Thus, during the quantization step 16, one of 15 possible quantizers R0 - R14 is selected and used for the coding of a frame or a portion of a frame, and the quantizer Rx may be changed for any frame or frame portion (i.e., 8 x 8 pel block or 16 x 16 pel block).
  • While the above example involves 15 quantizers R0 - R14, derived by generating all possible sets which differ at absolute values of non-zero QP multiples below 7, a greater number of quantizers may be generated for sets that differ at higher absolute values of QP multiples. Similarly, a lower number of quantizers may be generated for sets that differ at lower absolute values of QP multiples. Furthermore, by experimentation, the quantizer sets that have the highest probability of being selected from the overall group of quantizers may be determined and the number of possible quantizers may be limited to those high probability sets. For example, by experimentation with the above quantizers R0 - R14 and using a selection technique discussed below, it has been found that the quantizers R0, R3, R7 and R14 have the highest probability of selection.
  • Selection of the appropriate quantizer Rx may be carried out according to any suitable criterion and in the preferred embodiment of the present invention the selection is based on a formula which takes into account both distortion and bit rate, which provides the lowest error E: E = D + λr where D is a quantified value of the amount of distortion of the portion of the video image and r is a quantified bit rate which depends on the length of the bit code obtained in the coding step 18. Other formulas for calculating an error, for example, based only partially on the amount of distortion and bit rate factors, may also be used.
  • According to the above formula, the effect of the bit rate r on the overall error value E is weighted by a factor of λ. The value of λ may be predetermined as a design parameter, based on the desired degree of the effect of the bit-rate on the selection of quantizers, or computed in real-time as the function of the bit-rate. Alternatively, the value λr may be pre-computed for different overall bit-rates and stored for access by a processor to avoid calculating this product for each input sample. A preferred λ value of 2 has been found to provide superior results in many applications.
  • Thus, the quantization step 16 involves a calculation of the error value E, based on the amount of distortion and the bit rate provided by each quantizer in the selection group. Since this can result in a large number of calculations for systems employing a large number of quantizers, the quantizers may be limited to the quantizer sets that have been found to have the highest probability of selection, such as quantizer sets R0, R3, R7 and R14 in the above example.
  • An identification code, corresponding to the chosen Rx, must be provided to the receiver to identify which quantizer was selected and used during the quantization step 16 for each frame or frame portion. The number of bits required for the identification code is dependent on the number of quantizers in the selection group and for 15 quantizers defined above the identification code includes 4 bits, if the fixed-length code words are used to represent the identification code. If only four quantizers R0, R3, R7 and R14 are used, only two bits are needed. Alternatively, variable length code words may be used to represent quantizers (e.g., shorter length code words for representing more frequently chosen quantizers and longer length code words for less frequently chosen quantizers). Even with the inclusion of the quantizer identification code in the transmitted sequence for each coded frame or frame portion, an improvement in the overall bit rate, as well as a higher accuracy, was accomplished with the above described process over the standard methods.
  • Further improvements in bit rate and accuracy are provided by employing an error calculation step during the selection of a quantizer value (QP multiple) within a quantizer. In particular, instead of selecting the QP multiple which is the closest, or closest and not greater, to the calculated Y' value, as discussed above, an error calculation step is performed in which an error E' provided by each QP multiple is calculated and the quantizer value with the lowest error E' is selected as the estimated Y' value. As with the above-discussed error calculations, the error E' can be calculated from a formula which takes into account both the distortion and bit rate: E' = D' + λ'r' where D' is a quantified value for distortion amount and r' is a quantified value for bit rate. The value of λ' may be predetermined as a design parameter, based on the desired effect of the bit-rate on the overall error value E', as discussed above, it may be computed in real-time as the function of the bit-rate, or λ'r' may be pre-computed and stored. It has been found that a value of 60 for λ' provides superior results in many applications.
  • The selection of a specific quantizer among the group of quantizers that gives the best combined distortion and bit rate characteristics can provide significant improvements in the overall accuracy and bit rate of the coding system. Furthermore, the selection of a particular quantizer value (QP multiple) within the selected quantizer, which provides the best combined distortion and bit rate characteristics, can provide further improvements in the overall accuracy and bit rate of the coding system.
  • Such improvements are shown using standard ITU-T testing sequences in the following examples. In Example 1 the quantizer set R0 was used in accordance with the conventional H.263 standard. In Example 2 the quantizer group R0, R3, R7 and R14 was selected, and the selection was based on the above-discussed error equations and the indicated λ and λ' values. In Example 3 only the quantizer R0 was used, but the selection of quantizer values within it was based on the above-discussed error E' calculation. ITU-T "Carphone" Sequence (R=20kbps) Examples Example 1 Example 2 Example 3 λ and λ' none
    standard H.263
    λ=2
    λ'=60
    λ=0
    λ'=60
    distortion 30.77 dB 31.60 dB 31.04 dB bit rate 24.19 kbps 24.30 kbps 23.95 kbps frame rate 5.05 Hz 5.30 Hz 5.30 Hz no. of frames 20 frames 21 frames 21 frames
    ITU-T "Mother & Daughter" Sequence (R = 20 kbps) Examples Example 1 Example 2 Example 3 λ and λ' none
    standard H.263
    λ=2
    λ'=60
    λ=0
    λ'=60
    distortion 31.59 dB 32.08 dB 32.01 dB bit rate 23.47 kbps 21.97 kbps 23.17 kbps frame rate 5.25 Hz 7.25 Hz 5.50 Hz no. of frames 21 frames 29 frames 22 frames
    ITU-T "Susie" Sequence (R = 20 kbps) Examples Example 1 Example 2 Example 3 λ and λ' none
    standard H.263
    λ=2
    λ'=42
    λ=0
    λ'=42
    distortion 32.61 dB 33.48 dB 33.01 dB bit rate 22.95 kbps 23.26 kbps 22.94 kbps frame rate 5.10 Hz 5.41 Hz 5.21 Hz no. of frames 20 frames 21 frames 20 frames
  • As is apparent from each of the above examples, better distortion (signal to noise ratio), bit rate and frame rate characteristics were achieved in Examples 2 and 3, as compared to Example 1 which uses the standard quantizer R0. The examples show a gain in coding efficiency with an increase in the frame rate for the same quality of coding at the same bit rate, or an increase in quality and decrease in the bit rate for the same frame rate and a fixed quantization step size.
  • The present invention uses an adaptive entropy coding system and method in the coding step 18 of the adaptive quantization framework. In one preferred embodiment of the present invention a group of 3-D VLC tables is used for adaptive entropy coding (encoding and decoding) of each event, represented with a sequence (Level, Run, Last). The events correspond to pels with the highest probability of occurrence in the input signal and are stored in the VLC tables. Each 3-D VLC table also has an assigned bit code and a bit code sign for each event. The Level corresponds to the Level index to the quantizer corresponding to the quantizer value closest to the input signal pel value, the Run indicates number of zero valued pels before the current event and the non-zero value of Last indicates the last pel in a block which has non-zero value. VLC tables are determined by an off-line training procedure and each quantizer Rx has its own optimized VLC table in this preferred embodiment of the present invention.
  • In another preferred embodiment of the present invention only one, common, optimized VLC table is used for all quantizers. The lengths of the binary codes obtained in this embodiment, corresponding to the common VLC table, are presented in the tables of Fig. 6 and Fig. 7 for non-last and last events in a block.
  • Yet another preferred embodiment of the present invention is a coding processor which implements the adaptive entropy coding method of the present invention in the adaptive quantization framework. The coding processor could be a general computer or a special, dedicated computer. It is used to perform the encoding of a macroblock, which includes a determination of the quantizer Rx closest to the input video signal in the quantization step 16. Once the quantizer is determined, the Level index for each quantizer value of each event is obtained as follows.
  • To get the actual quantizer values for a block of 8*8 pels, the most important, DC signal pel value is first considered. In the block with values represented in frequency domain, obtained in the DCT step 12, it is the top left element of the 8*8 block. At the encoder side, the Level for this DC pel is generated by the equation Level=actual value of DC pel/8. Hence, on the decoder side, a reconstructed signal value for the DC pel is calculated as Rec = Level*8. For all other video signal input values, in the encoding process the Level index for each pel is obtained during the quantization step 16 using the above-mentioned equation for calculation of the error E', which takes into account both the distortion and bit rate: E' = D' + λ'r'
  • The Level index represents the index to the one-dimensional Quantizer Value Array of the quantizer Rx, whose elements have the closest match to the actual pel values, together with the sign of the pel values. For example, if the pel value is 45 and the chosen quantizer is R2, QP=16 and α =1, than the Quantizer R2 can be represented by the equation: R2 = {0, (3*16-1), (4*16-1), 7*16-1), ...., (255*16-1)}= {0,47, 63, 111,..., 4079}
  • In this case the Level index Level=1, because the input pel value of 45 is the closest to the first non-zero quantizer value 47 in the quantizer R2.
  • In the coding step 18, Level=1 is coded together with the Run and Last value and the corresponding binary code, selected from the VLC table, is sent to the decoder together with the sign of the actual pel input value. In addition, for each macroblock a fixed-length binary identification code is transmitted to the decoder to indicate which of the quantizers is being used for decoding of the macroblock.
  • In the decoder, the first step consists of determining the used quantizer Rx (here R2) by decoding the fixed-length code. Then the actual reconstructed quantizer value of each pel is recovered by decoding the binary code of each event (Level, Run, Last) to get the quantizer value (here 47) from the quantizer R2, which corresponds to the coded Level index (here 1), using the same 3-D VLC table corresponding to the quantizer used in the encoder side. The actual reconstructed signal value at the decoder is calculated as Rec = |Rec| * sign(Level). |Rec| is obtained from the reconstructing equation: |Rec| = R_{I}[|Level| ] where R_{I}[.] represents the chosen quantizer, Level is the Level index, and |Rec| is the magnitude of the reconstructed value which corresponds to the quantizer value 47 in this example.
  • In another preferred embodiment of the present invention, the quantization step 16 is performed after an offset γ is given to each pel of a macroblock. Each pel value x; is biased towards a "0" level by the offset. This further improves the matching of the rate-distortion characteristics of the macroblocks to the chosen quantizer Rx. The method of performing the quantization step 16 using an offset γ is illustrated in the operational block diagram of Fig. 11. Initially in step 100, a quantizer Rx is selected from the group of n quantizers for a particular macroblock of pel values. A macroblock is a group of several blocks of pels, such as a group of several 8x8 blocks of pels. A variable offset γ is then selected in step 102 for each block, where the offset γ is within a predetermined range -t < γ < t. The pel values xi for a block are then modified by the offset γ in step 104 to form modified pel values (xi - γ). Each modified pel value in the block is then mapped to the closest available quantizer value yi from the chosen quantizer Rx in step 106, where Rx = {0, ±y1, ±y2, ... ±yz-1} for z number of quantizer levels. After quantizing the entire block, an effective bit rate reff incurred in quantizing this block with the selected quantizer Rx and the selected offset γ is calculated in step 108. The effective bit rate reff can be determined in any known manner, such as being accurately calculated by a table look-up operation within the actual Huffman tables used in transmitting the levels and runs of the block. The overall distortion D for the block is determined in step 110 by summing the distortion over all of the pels as follows:
    Figure 00190001
    where n is the number pels in the selected block.
  • A weighted distortion E which takes into account both the distortion D and effective bit rate reff for the entire block is then determined in step 112 for the chosen offset γ according to the equation: E = D + λreff where λ is a predetermined constant. The offset γ is then varied over the range -t < γ < t in step 114, and steps 102 through 112 are repeated for the same block using the same quantizer Rx for various values in the range -t < γ < t. The value of the offset γ found to minimize the weighted distortion E for the particular block analyzed is determined to be the optimal offset for that block for the selected quantizer Rx in step 116. Using the same selected quantizer Rx, steps 102 through 116 are repeated for each of the remaining blocks in the macroblock in step 118. An overall weighted distortion Eall for the entire macroblock is determined in step 120 by summing the least weighted distortion E obtained in step 116 for each individual block. In step 122, the adaptive quantization scheme selects a different quantizer Rx from the group of n quantizers and repeats steps 100 through 120 to determine the least overall weighted distortion Eall for the newly selected quantizer Rx. This procedure is repeated until a least overall weighted distortion Eall is obtained for each quantizer in the group of n quantizers. For each quantizer, one particular offset γ will provide the least overall weighted distortion Eall using that particular quantizer. Finally, in step 124, the quantizer which minimizes the overall weighted distortion Eall for the entire macroblock along with an associated offset is selected as the optimal quantizer Ropt.
  • In yet another preferred embodiment of the present invention, the quantization step 16 is performed by searching through a trellis of decision levels which correspond to all possible choices of level values yi and run values of consecutive zeros Zi, as illustrated in FIG. 12. The trellis includes a plurality of stages equal to the number of pels being quantized. Each pel is mapped to a corresponding decision level, either a level value yi or zero run value Zi. The transition from a decision level in the trellis at one stage to another decision level in the next stage is weighted by an associated distortion and bit rate. In this embodiment, the distortion choices and corresponding bit rates from the transitions are analyzed to determine an optimal trellis path for all of the pels in a block which provides a minimum overall weighted distortion in order to further improve the matching of the rate-distortion characteristics of the macroblocks to a chosen quantizer Rx.
  • The optimal trellis path is selected using a Viterbi algorithm or other similar algorithm on the trellis. The particular trellis illustrated in FIG. 12 is set up for an 8x8 block of 64 pels, resulting in 64 stages in the trellis search. The arrows linking the values between stages illustrate possible transitions that can be made from stage to the next, where FIG. 12 only shows a few selected possible transitions for illustrative purposes and does not show each and every possible transition between the various stages. There are certain guidelines which govern the possible transitions that can be made. For instance, a zero run state Zi+1 can only be reached from a zero run state Zi in the previous stage. Furthermore, the initial zero run state Z1 can only be reached from a level value state yi in the previous stage. Once a zero run state Zi ends with a level value state yi or once the last stage of the trellis is reached, the trellis path can be traced back to determine the distortion and rate associated with that particular path. It is understood to those skilled in the art that the particular trellis utilized will vary depending upon the number of pels in the block and the number of quantization levels in the quantizer being analyzed.
  • The method of performing the quantization step 16 using an optimal trellis path is illustrated in the operational block diagram of Fig. 13. Initially in step 200, a quantizer Rx is selected from the group of n quantizers for a particular macroblock. For each block in the macroblock, the decision as to which level to map a pel value to is determined by traversing the trellis through all of the possible choices of level values yi and consecutive zero run values Zi for the selected quantizer Rx in step 202, where the level values yi for selected quantizer Rx are represented by the set Rx ={0, ±y1, ±y2, ... ±yz-1} for z number of quantizer levels. The trellis is analyzed to determine the amount of distortion and corresponding bit rates ri for each transition through the stages in the trellis. The bit rate ri is the rate of the corresponding level yi taking into account the preceding zero run-lengths. The rates ri may be predetermined and stored for the selected quantizer Rx, such as in a Huffman table. Each transition has an associated weighted distortion value E calculated according to the formula: E=D + λri, where D is the distortion value between decision levels from one stage to the following stage and λ is a weighting value. A weighted distortion E' for the trellis is determined by summing the weighted distortion values E through all of the stages of the trellis. The quantizer mapping for all of the pels in a given block is determined by selecting the path in the trellis providing the least overall weighted distortion EL in step 204. The trellis path may be determined using a Viterbi algorithm or any other dynamic programming algorithm. This least overall weighted distortion EL provides the most optimal rate-distortion tradeoff for the quantizer selected for the given block. The trellis path providing the least weighted distortion EL for the given block is then applied to the remaining blocks in the macroblock to obtain an overall weighted distortion Eall for the entire macroblock in step 206.
  • The trellis path providing the least overall weighted distortion Eall for the entire macroblock is determined in step 208 to provide the most optimal rate-distortion tradeoff for the selected quantizer Rx. In step 210, the adaptive quantization scheme selects a different quantizer Rx from the group of n quantizers and repeats steps 200 through 208 to determine the trellis path providing least overall weighted distortion Eall for the macroblock for the newly selected quantizer Rx. This procedure is repeated for each of the possible quantizers until the quantizer providing the least overall weighted distortion Eall for the given macroblock is selected as the optimal quantizer Ropt along with its associated optimal trellis path in step 212.
  • In the preferred embodiment of the present invention a training procedure is used on training signals, similar to the video signals to be used with the coding system, to obtain the VLC table, later used with the actual video signals. During the training procedure, adaptive quantization is used to find the best matching quantizer which produces the smallest error, and a count of occurrences for each event is kept in a counter table corresponding to the chosen quantizer. At the end of the training procedure the probability of occurrence of each event (Level, Run, Last) in the training signal is calculated. These probabilities are then used to generate a probability table, which has probabilities for each combination of Level, Run and Last. The information from the probability table is then used to define a corresponding binary code for the combination of Level, Run and Last for each event, and the binary code are stored in a VLC table, so that the shortest binary code is assigned to the event which has the highest probability of occurrence. In the preferred embodiment which uses a separate VLC table for each quantizer, there are also multiple counter tables and probability tables, one of each for each quantizer.
  • The probability is calculated from the total number of occurrences of an event in the training signal. For the events that do not have high probability of occurrence, an Escape (Esc) code is entered in the tables. For less probable events, instead of defining coded values for Level, Run and Last and storing them in the VLC table, fixed-length codes for actual values for Level and Run of that event are sent to the decoder. This makes the procedure more efficient because it keeps the VLC table short. Since each Run can have a value between 0 and 63, it can be represented with six bits (26=64). Since each Level can have a value between -127 and +127, it can be represented with eight bits (28=256). Therefore, fixed-length codes for an event are 14 bits long, which is much longer than the average length of binary codes chosen from the VLC table. When an Esc code is encountered in the decoder, it is treated as an actual value for Run and Level until another Esc code is received.
  • The encoding of the actual video signals is accomplished in the variable length coding step 18 with the help of the VLC table defined during the training procedure. The pel values of the block are coded in the coding processor with a suitable entropy coding scheme such as Huffman Coding, Arithmetic Coding, Syntax Adaptive Arithmetic Coding or Ziv-Lempel Coding. Huffman coding is preferred because of its cost, accuracy and efficiency, because pel values can be coded with a relatively short binary code sequence, and because the number of zeroes in strings of zeros, which tend to occur toward the lower right corner of the quantized block, is included in the code for the next pel and these events do not need separate code. The efficiency of this method can be enhanced by coding the data for each pel in the zig-zag sequence, as mentioned above.
  • The method of the present invention utilizes joint probability distribution of three event parameters, Level, Run and Last, for each event, which further reduces the bit rate by reducing the average binary code length for each event. Since the binary codes for all events of a sequence are concatenated to form a stream of binary digits for storage or transmission, the correct decoding by the receiver requires that every combination of concatenated codes be uniquely decipherable. A sufficient condition is that the binary code satisfies the prefix rule, which requires that no binary code word may be the prefix of any other code word in the VLC table. Huffman code satisfies the prefix rule and it is very efficient because it performs within one bit/pel of the entropy. However, the present invention can also create each VLC table with a different entropy coding method, with or without Huffman's code.
  • The two tables presented in Figs. 6 and 7 show examples of the lengths of binary codes in the 3-D VLC table obtained by the training method of the present invention, when only one common VLC table is used for all quantizers. Each actual VLC code also contains a 1-bit sign value added at the end of the binary code for the combination of Level, Run and Last. The tables of Figs. 6 and 7 have limited numbers of Runs and Levels; these numbers are determined by probabilities of the events and are kept as small as possible according to the accuracy needed. Therefore, in the table of Fig. 6 for events that are not last, twelve Level indexes and twenty-seven Runs are sufficient. In the table of Fig. 7 for last events, only three level indexes and forty-one Runs are needed, instead of 256 Level indexes and 64 Runs. The tables shown in Figs. 6 and 7 are presented for exemplary purposes and are not intended to represent all possible tables which may utilized by the present invention. It is understood that other possible tables may be generated by the present invention.
  • In an exemplary use of the method of the present invention, an 8x8 pel input block, given in Fig. 8, was quantized using the adaptive quantization scheme and quantizer R2 was chosen from the group of 15 quantizers. The quantizer values for R2 are defined as: R2 = {0,3*QP-α, 4*QP-α, 7*QP-α, ....,255*QP-α} where α = 1 if QP is an even number and otherwise is 0.
  • Although the quantization step size QP can change during the coding process for each portion of a block, as decided by the computer program with the coding algorithm, it is assumed in this example that for this particular block QP = 16 and hence α =1. The quantizer values are therefore: R2 = {0, (3*16-1), (4*16-1), 7*16-1), ...., (255*16-1)}= {0, 47, 63, 111, ..., 4079}
  • In quantizing the input block of Fig. 8 with the quantizer R2, each input pel value is matched to the nearest quantizer value of the quantizer R2 and the Level index of each matched quantizer value is saved in a quantized block which corresponds to R2, shown in Fig. 9. This quantized block is then linearly scanned to get the event (Level, Run, Last) for each pel of this block, preferably in the zig-zag fashion, to obtain a block represented in Fig. 10. The more sophisticated method to obtain Level would be to use the above-mentioned equation for calculation of the error E' which takes into account both distortion and bit rate: E' = D' + λ'r'
  • From the block represented in Fig. 10, it can be seen that there are only five non-zero events Event_1 to Event_5:
  • Event_1 = [Level=3, Run=0, Last=0]
  • Event_2 = [Level=1, Runs=0, Last=0]
  • Event_3 = [Level=2, Run=0, Last=0]
  • Event_4 = [Level=2, Run=1, Last=0]
  • Event_5 = [Level=1, Run=0, Last=1]
  • Run =1 in Event_4 shows that there is 1 zero before this event. Event_5 has Last=1 to show that this is the last non-zero event.
  • If this input block was the training block used during the training procedure, the counts for these five events in the counter table for quantizer R2 will be incremented to show that five events had input pel values that may be quantized with R2. By extending the training process over several training sequences, different quantizers will be chosen for different blocks or portions of blocks of the training sequence. For each of the blocks, according to the chosen quantizer, the appropriate counter table for each event will be updated and used in calculation of probability of the event. Since these probabilities are directly related to the binary codes in the VLC table, a different VLC table may be created for each quantizer. Because each binary code length corresponds to the probability that the particular event is quantizable with that quantizer, use of a separate VLC table for each quantizer produces binary codes with optimized length.
  • In the above example, a binary code for each of the 5 events is transmitted to the receiver for encoding, which also receives a 4-bit code for the quantizer used for the whole macroblock, in this example R2. Therefore, the receiver selects the VLC binary code table corresponding to the quantizer used in the encoder in order to decode the Level index, Run and Late from the transmitted binary code.
  • The tables in the training process may be obtained in several different ways. One method of training and generating the probabilities, and eventually the VLC binary code tables, is to set the quantization step size QP to a fixed value used throughout the training process and to use the probabilities of different events, obtained for different quantizers, to generate the VLC tables for all quantizers. In another training method, the probabilities of different events for different quantizers are obtained by training at different QPs and the probabilities for each event at different QPs are averaged. Therefore, the training procedure may be chosen according to the input signal type. However, regardless of what the training procedure is, if the conditions during training are general enough, VLC tables may be obtained which are robust enough for a wide variety of environments.
  • Once the VLC tables for each of the quantizers are made in the training process, the choice of the appropriate VLC table may be made by the same distortion criteria as used for choosing the quantizer, with the distortion equation E = D +λ *r, which is calculated for each pel of the block, when the quantizer which results in the least cumulative weighted distortion for the whole macroblock is chosen. However, the choice of a quantizer, and hence the VLC table associated with that quantizer, can be made by using a different distortion equation. For example, by incorporating some other parameters, like QP, the distortion equation may be defined as: E = D + λ*f(QP)*r where f(QP) is a function of QP, like for example f(QP)=K/QP, where K is a constant determined by experimentation.
  • In one embodiment of the present invention a different VLC table is defined for each of the 15 quantizers. By having a separate VLC table for each quantizer an even more efficient compression of Levels and Runs can be accomplished by coding. With the method of the present invention a 10-25% gain in bit rates is achieved, which is considered substantial in the video coding field.
  • The adaptive entropy coding in adaptive quantization coding algorithm and the system of the present invention are believed to be especially applicable and useful in circuits used in low bit rate visual telephony for video conferencing, video codecs, set top boxes and the like. It is understood that the principles of this invention may also be applied in circuits used in non-adaptive environments and for higher bit rate video coding.

Claims (34)

  1. An adaptive quantization method in a coding process for coding data values corresponding to each picture element (pel) of a portion of a video frame, the method comprising the steps of:
    selecting, for each portion of a video frame, a quantizer set Rx of quantizer values from a group of plurality of quantizer sets R0 - Rn, where each quantizer set differs from each other quantizer set in the group, said quantizer values being saved in a quantizer value array corresponding to said quantizer set Rx;
    wherein the selecting comprises the steps of:
    determining the distortion rate and bit rate characteristics provided by each quantizer set the group R0 - Rn; and
    selecting the quantizer set Rx based on the distortion and bit rate characteristics provided by each quantizer set in the group; and
    correlating each data value with a quantizer value from the selected quantizer set Rx to provide a quantized set of values corresponding to the pel data values.
  2. The method as recited in Claim 1, wherein said step of selecting comprises the steps of:
    determining the distortion and bit rate characteristics provided by each quantizer set in the group R0 - Rn; and
    selecting the quantizer set Rx based on the distortion and bit rate characteristics provided by each quantizer set in the group.
  3. The method as recited in Claim 1, wherein said step of correlating comprises the steps of:
    determining the distortion and bit rate characteristics provided by each quantizer value in the selected quantizer set Rx; and
    selecting the quantizer value based on the distortion and bit rate characteristics provided by each quantizer set in the group.
  4. The method as recited in Claim 1, wherein said step of correlating comprises the steps of:
    determining and quantifying the distortion and bit rate characteristics provided by each quantizer value in the selected quantizer set Rx;
    calculating an error value E' for each pel and for each quantizer value in the set Rx according to the formula: E' = D + λ'r', where D' is a quantified distortion value, r' is a quantified bit rate value and λ' is a weighting value selected based on the desired effect of r' on the overall E'; and
    selecting the quantizer value that exhibits the lowest E' value of the quantizer values within the set Rx.
  5. The method as recited in Claim 1, further comprising the step of providing an identification code representing the quantizer Rx selected from the group R0-Rn.
  6. The method as recited in Claim 1, wherein said step of selecting comprises the steps of:
    determining and quantifying the distortion and bit rate characteristics provided by each quantizer set in the group R0 - Rn for each portion of the video frame being coded;
    calculating an error value E for each quantizer set in the group R0 - Rn according to the formula: E = D + λr, where D is a quantified distortion value, r is a quantified bit rate value and λ is a weighting value selected based on the desired effect of r on the overall E; and
    selecting the quantizer set Rx that exhibits the lowest E value of the quantizer sets within the group R0 - Rn.
  7. The method as recited in Claim 6, wherein λ = 2.
  8. The method as recited in Claim 6, wherein said step of correlating comprises the steps of:
    determining the distortion and bit rate characteristics provided by each quantizer value in the selected quantizer set Rx; and
    selecting the quantizer value based on the distortion and bit rate characteristics provided by each quantizer value in the set Rx.
  9. The method as recited in Claim 6, wherein said step of correlating comprises the steps of:
    determining and quantifying the distortion and bit rate characteristics provided by each quantizer value in the selected quantizer set Rx;
    calculating an error value E' for each quantizer value in the set Rx according to the formula: E' = D' + λ'r', where D' is a quantified distortion value, r' is a quantified bit rate value and λ' is a weighting value selected based on the desired effect of r' on the overall E'; and
    selecting the quantizer value that exhibits the lowest E' value of the quantizer values within the set Rx.
  10. The method as recited in Claim 1, wherein each quantizer value in each set within the group R0 - Rn is a multiple of a quantization coefficient QP.
  11. The method as recited in Claim 10, wherein the quantizer sets within the group R0 - Rn are selected from the group consisting of: R0 = {0,±3QP-α,±5QP-α,±7QP-α ...±255QP-α} R1 = {0,±3QP-α,±6QP-α,±7QP-α ... ±255QP-α} R2 = {0,±3QP-α,±4QP-α,±7QP-α, ... ±255QP-α} R3 = {0,±2QP-α,±6QP-α,±7QP-α, ... ±255QP-α} R4 = {0,±2QP-α,±5QP-α,±7QP-α, ... ±255QP-α} R5 = {0,±2QP-α,±4QP-α,±7QP-α, ... ±255QP-α} R6 = {0,±2QP-α,±3QP-α ±7QP-α, ... ±255QP-α} R7 = {0,±4QP-α,±6QP-α,±7QP-α,...±255QP-α} R8 = {0,±4QP-α,±5QP-α,±7QP-α,...±255QP-α} R9 = {0,±1QP-α,±6QP-α,±7QP-α,...±255QP-α} R10 = {0,±1QP-α,±5QP-α,±7QP-α...±255QP-α} R11 = {0,±1QP-α,±4QP-α ±7QP-α,...±255QP-α} R12 = {0,±1QP-α,±3QP-α,±7QP-α,...±255QP-α} R13 = {0,±1QP-α,±2QP-α,±7QP-α,...±255QP-α} R14 = {0,±5QP-α,±6QP-, α ±7QP-α,...±255QP-α}    where the quantization step size QP can be any number between 1 and 31 and α = 1 if QP is an even number and α = 0 if QP is an odd number.
  12. The method as recited in Claim 10, wherein the quantizer sets within the group R0 - Rn are selected from the group consisting of: R0 = {0,±3QP-α,±5QP-α,±7QP-α,...±255QP-α}; R3 = {0,±2QP-α,±6QP-α,±7QP-α,...±255QP-α}; R7 = {0,±4QP-α,±6QP-α,±7QP-α,...±255QP-α}; and R14 = {0,±5QP-α,±6QP-α,±7QP-α,...±255QP-α}.    where the quantization step size QP can be any number between 1 and 31 and α = I if QP is an even number and α = 0 if QP is an odd number.
  13. The method as recited in Claim 1, wherein said quantizer values are multiples of a quantization coefficient QP, comprising the steps of:
    selecting at least one VLC table having an entry for each pel event with a high probability of occurrence; and
    coding the quantizer values with variable length adaptive entropy coding method using the at least one VLC table.
  14. The method as recited in Claim 13, wherein said step of adaptively quantizing the sampled data values comprises the steps of:
    determining and quantifying the distortion and bit rate characteristics provided by each quantizer set in the group R0 - Rn for each portion of the video frame being coded;
    calculating an error value E for each quantizer set in the group R0 - Rn according to the formula: E = D + λ f(QP)r, where D is a quantified distortion value for the portion of the video frame, r is a quantified bit rate value for the portion of the video frame, λ is a weighting value selected based on the desired effect of r on the overall E and f(QP) is a function of the quantization coefficient QP;
    selecting the quantizer set Rx that exhibits the lowest E value of the quantizer sets within the group Ro - Rn;
    calculating an error value E' for each pel of said portion of the video frame and for each quantizer value in the quantizer set Rx according to the formula: E' = D' + λ'r', where D' is a quantified distortion value for the pel, r' is a quantified bit rate value for the pel and λ' is a weighting value selected based on the desired effect of r' on the overall E'; and
    selecting the quantizer value that exhibits the lowest E' value for the pel.
  15. The method as recited in Claim 13, wherein said variable length adaptive entropy coding method is chosen from a group comprised of Huffman Coding, Arithmetic Coding, Syntax Adaptive Arithmetic Coding and Ziv-Lempel Coding method, so that the chosen binary code satisfies the prefix rule which requires that no binary code word may be a prefix of any other binary code word in the VLC table.
  16. The method as recited in Claim 13, wherein:
    each said entry in the VLC table is the event defined with three event parameters, a Level, a Run, and a Last;
    the Level is a level index to the quantizer set Rx corresponding to the selected quantizer value;
    the Run indicates a number of zero pels before the current event, which is included in the binary code for the current event; and
    the Last indicates the last pel in a block which has non-zero value.
  17. The method as recited in Claim 16, wherein said coding step comprises a joint probability distribution binary coding step for simultaneous coding of all three event parameters, the Level, the Run and the Last, to reduce an average binary code length for each event, and wherein
       the binary code for each event is saved with the sign of the actual pel input value in the VLC table.
  18. The method as recited in Claim 17, wherein the step selecting the at least one VLC table further comprising a training step performed on a training signal, comprising the following steps:
    keeping a count of occurrences for each event from the training signal in at least one counter table;
    calculating a probability of occurrence of each said event from the total number of occurrences of the event in the training signal;
    generating at least one probability table to save probabilities for each combination of the parameters Level, Run and Last, to be used in the coding step, so that the shortest binary code is assigned to the event which has the highest probability of occurrence; and
    using an adaptive quantization step to find the best matching quantizer for each said event which produces the smallest error.
  19. The method as recited in Claim 18, wherein the training step is calculating probabilities for each said event using each quantizer set Rx with all possible values of the quantization coefficient QP and the probabilities for each said event at different QPs are averaged.
  20. The method as recited in Claim 18, wherein the training step generates a plurality of VLC tables, a plurality of counter tables and a plurality of probability tables, one of each for each quantizer Rx.
  21. The method as recited in Claim 18, wherein the training step generates one common optimized VLC table used for all quantizers.
  22. The method as recited in Claim 13, wherein the quantizer sets within the group R0 - Rn are selected from the group consisting of: R0 = {0,±3QP-α,±5QP-α,±7QP-α...±255QP-α} R1 = {0,±3QP-α,±6QP-α,±7QF-α... ±255QP-α} R2 = {0,±3QP-α,±4QP-α,±7QP-α,...±255QP-α} R3 = {0,±2QP-α,±6QP-α,±7QP-α,...±255QP- α} R4={0,±2QP-α,±5QP-α,±7QP-α,...±255QP-α} R5={0,±2QP-α,±4QP-α,±7QP-α,...±255QP-α} R6 ={0,±2QP-α,±3QP-α±7QP-α,...±255QP-α} R7 = {0,±4QP-α,±6QP-α,±7QP-α,...±255QP-α} R8 = {0,±4QP-α,±5QP-α,±7QP-α,...±255QP-α} R9 = {0,±1QP-α,±6QP-α,±7QP-α,...±255QP-α} R10 = {0,±1QP-α,±5QP-α,±7QP-α...±255QP-α} R11 = {0,±1QP-α,±4QP-α ±7QP-α, ... ±255QP-α} R12 = {0,±1QP-α,±3QP-α,±7QP-α,...±255QP-α} R13 = {0,±1QP-α,±2QP-α,±7QP-α,...±255QP-α} R14 = {0,±5QP-α,±6QP-,α ±7QP-α,...±255QP-α}    where the quantization coefficient QP is an integer between 1 and 31, and α = 1 if QP is an even number and α = 0 if QP is an odd number.
  23. The method as recited in Claim 13, wherein the quantizer sets within the group R0 - Rn are selected from the group consisting of: R0 = {0, ±3QP-α,±5QP-α,±7QP-α,...±255QP-α}; R3 = {0,±2QP-α,±6QP-α,±7QP-α,...±255QP-α}; R7 = {0,±4QP-α,±6QP-α,±7QP-α,...±55QP-α}; and R14 = {0,±5QP,±6QP,±7QP,...±255QP}.    where the quantization coefficient QP is an integer between 1 and 31, and α = 1 if QP is an even number and α = 0 if QP is an odd number.
  24. A system for coding sampled video signal data corresponding to picture elements (pels) of at least a portion of a video frame, comprising:
    a storage means for storing a group of a plurality of quantizer sets R0 - Rn, where each quantizer set differs each other quantizer set in the group; and
    a processor means comprising:
    means for selecting, for each portion of the video frame being coded, a quantizer set Rx, of quantizer values, which are multiples of a quantization coefficient QP, from a group of sets R0 - Rn, dependent of a quantified distortion and bit rate of the selected set Rx determined for each said portion of said video frame;
    means for correlating each video signal data value for each frame portion being coded with a quantizer value from the Rx selected for that frame portion;
    means for selecting at least one VLC table having an entry for each pel event with a high probability of occurrence; and
    means for encoding the quantizer values with a variable length adaptive entropy encoding method using the at least one VLC table.
  25. The system as recited in Claim 24, wherein said selecting means comprises:
    means for calculating an error value E for each portion of the video frame being coded and for each quantizer set in the group Ro - Rn, according to the formula: E = D + λ f(QP)r, where D is a quantified distortion value for the portion of the video frame, r is a quantified bit rate value for the portion of the video frame, λ is a weighting value selected based on the desired effect of r on the overall E and f(QP) is a function of the quantization coefficient QP; and
    means for selecting the quantizer set Rx that exhibits the lowest E value of the quantizer sets within the group Ro - Rn; and
       wherein said correlating means comprises:
    means for determining and quantifying the distortion and bit rate characteristics of each quantizer value within the set Rx; and
    means for selecting the quantizer value based on the distortion and bit rate characteristics provided by each quantizer value in the set Rx.
  26. The system as recited in Claim 24, wherein said variable length adaptive entropy coding method is chosen from a group comprised of Huffman Coding, Arithmetic Coding, Syntax Adaptive Arithmetic Coding and Ziv-Lempel Coding method, so that the chosen binary code satisfies the prefix rule which requires that no binary code word may be a prefix of any other binary code word in the VLC table.
  27. The system as recited in Claim 24, wherein:
    each said entry in the VLC table is the event defined with three event parameters, a Level, a Run, and a Last;
    the Level is a level index to the quantizer set Rx corresponding to the selected quantizer value;
    the Run indicates a number of zero pels before the current event, which is included in the binary code for the current event; and
    the Last indicates the last pel in a block which has non-zero value.
  28. The system as recited in Claim 27, wherein said variable length adaptive entropy coding method comprises a joint probability distribution binary coding method for simultaneous coding of all three event parameters, the Level, the Run and the Last, to reduce an average binary code length for each event, and wherein
       the binary code for each event is saved with the sign of the actual pel input value in the VLC table.
  29. The system as recited in Claim 27, wherein the at least one VLC table is determined in a training procedure with the processor means using a training signal, said processor means further comprising:
    means for keeping a count of occurrences for each event from the training signal in at least one counter table;
    means for calculating a probability of occurrence of each said event from the total number of occurrences of the event in the training signal; and
    means for generating at least one probability table to save probabilities for each combination of the parameters Level, Run and Last, to be used in the encoding means so that the shortest binary code is assigned to the event which has the highest probability of occurrence.
  30. The system as recited in Claim 29, wherein the training procedure is calculating probabilities for each said event using each quantizer set Rx with all possible values of the quantization coefficient QP and the probabilities for each said event at different QPs are averaged.
  31. The system as recited in Claim 29, wherein the training procedure generates a plurality of VLC tables, a plurality of counter tables and a plurality of probability tables, one of each for each quantizer Rx.
  32. The system as recited in Claim 29, wherein the training procedure generates one common optimized VLC table used for all quantizers wherein length of the binary codes from the VLC table are represented in the following tables 6 and 7: L=1 L=2 L=3 1=4 L=5 L=6 L=7 L=8 L=9 10 11 12 R=1 1 3 5 5 6 7 8 8 9 9 11 11 2 2 6 7 8 10 10 9 10 10 11 Esc Esc 3 3 8 9 11 15 Esc 15 Esc Esc Esc Esc Esc 4 4 8 10 12 11 15 15 Esc Esc Esc Esc Esc 5 4 9 10 Esc 13 Esc Esc Esc Esc Esc Esc Esc 6 5 9 12 13 Esc Esc Esc Esc Esc Esc Esc Esc 7 5 11 14 Esc Esc Esc Esc Esc Esc Esc Esc Esc 8 5 9 10 15 Esc Esc Esc Esc Esc Esc Esc Esc 9 5 11 13 12 Esc Esc Esc Esc Esc Esc Esc Esc 10 6 15 15 14 Esc Esc Esc Esc Esc Esc Esc Esc 11 6 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc 12 7 12 14 Esc Esc Esc Esc Esc Esc Esc Esc Esc 13 7 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc 14 7 14 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc 15 8 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc 16 8 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc 17 9 15 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc 18 9 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc 19 9 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc 20 9 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc 21 10 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc 22 10 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc 23 12 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc 24 10 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc 25 10 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc 26 11 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc 27 12 Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc Esc
    LEVEL = 1 LEVEL = 2 LEVEL = 3 RUN = 1 2 7 6 RUN = 2 3 8 7 RUN = 3 3 9 13 RUN = 4 4 13 12 RUN = 5 4 Esc 12 RUN = 6 4 Esc 13 RUN = 7 4 Esc Esc RUN = 8 4 Esc Esc RUN = 9 5 Esc Esc RUN = 10 5 Esc Esc RUN = 11 5 Esc Esc RUN = 12 5 11 Esc RUN = 13 5 13 Esc RUN = 14 6 Esc Esc RUN = 15 6 Esc Esc RUN = 16 6 Esc Esc RUN = 17 6 Esc Esc RUN=18 7 Esc Esc RUN = 19 6 Esc Esc RUN=20 7 Esc Esc RUN=21 8 Esc Esc RUN = 22 7 Esc Esc RUN = 23 8 Esc Esc RUN=24 7 Esc Esc RUN=25 7 Esc Esc RUN = 26 7 Esc Esc RUN = 27 8 Esc Esc RUN = 28 8 Esc Esc RUN = 29 8 Esc Esc RUN = 30 9 Esc Esc RUN = 31 8 Esc Esc RUN = 32 9 Esc Esc RUN = 33 9 Esc Esc RUN = 34 8 Esc Esc RUN = 35 10 Esc Esc RUN = 36 9 Esc Esc RUN = 37 10 Esc Esc RUN = 38 9 Esc Esc RUN = 39 9 Esc Esc RUN = 40 8 Esc Esc RUN = 41 9 Esc Esc
  33. The system as recited in Claim 24, wherein the quantizer sets within the group R0 - Rn are selected from the group consisting of: R0 = {0,±3QP-α,±5QP-α,±7QP-α ... ±255QP-α} R1 = {0,±3QP-α,±6QP-α,±7QP-α ... ±255QP-α} R2 = {0,±3QP-α,±4QP-α,±7QP-α,...±255QP-α} R3 = {0,±2QP-α,±6QP-α,±7QP-α,...±255QP-α} R4 = {0,±2QP-α,±5QP-α,±7QP-α,...±255QP-α} R5 = {0,±2QP-α,±4QP-α,±7QP-α,...±255QP-α} R6 = {0,±2QP-α,±3QP-α ±7QP-α,...±255QP-α} R7 = {0,±4QP-α,±6QP-α,±7QP-α,...±255QP-α} R8 = {0,±4QP-α,±5QP-α,±7QP-α,...±255QP-α} R9 = {0,±1QP-α,±6QP-α,±7QP-α,...±255QP-α} R10 = {0,±1QP-α,±5QP-α,±7QP-α...±255QP-α} R11 = {0,±1QP-α,±4QP-α±7QP-α,...±255QP-α} R12 = {0,±1QP-α,±3QP-α,±7QP-α,...±255QP-α} R13 = {0,±1QP-α,±2QP-α,±7QP-α,...±255QP-α} R14 = {0,±5QP-α,±6QP-, a ±7QP-α,...±255QP-α}    where the quantization coefficient QP is an integer between 1 and 31, and α = 1 if QP is an even number and α = 0 if QP is an odd number.
  34. The system as recited in Claim 24, wherein the quantizer sets within the group R0 - Rn are selected from the group consisting of: R0 = {0, ±3QP-α, ±5QP-α, ±7QP-α, ... ±255QP-α}; R3 = {0, ±2QP-α, ±6QP-α, ±7QP-α, ... ±255QP-α}; R7 = {0, ±4QP-α, ±6QP-α, ±7QP-α, ... ±255QP-α}; and R14 = {0, ±5QP-α, ±6QP-α, ±7QP-α,... ±255QP-α}.    where the quantization coefficient QP is an integer between 1 and 31, and α = 1 if QP is an even number and α = 0 if QP is an odd number.
EP19980960467 1997-12-01 1998-11-25 Adaptive entropy coding in adaptive quantization framework for video signal coding systems and processes Not-in-force EP1033036B1 (en)

Priority Applications (5)

Application Number Priority Date Filing Date Title
US08/980,654 US6111914A (en) 1997-12-01 1997-12-01 Adaptive entropy coding in adaptive quantization framework for video signal coding systems and processes
US980654 1997-12-01
US09/198,742 US6118822A (en) 1997-12-01 1998-11-24 Adaptive entropy coding in adaptive quantization framework for video signal coding systems and processes
PCT/US1998/025155 WO1999029115A1 (en) 1997-12-01 1998-11-25 Adaptive entropy coding in adaptive quantization framework for video signal coding systems and processes
US198742 2002-07-19

Publications (2)

Publication Number Publication Date
EP1033036A1 EP1033036A1 (en) 2000-09-06
EP1033036B1 true EP1033036B1 (en) 2003-02-12

Family

ID=26894110

Family Applications (1)

Application Number Title Priority Date Filing Date
EP19980960467 Not-in-force EP1033036B1 (en) 1997-12-01 1998-11-25 Adaptive entropy coding in adaptive quantization framework for video signal coding systems and processes

Country Status (7)

Country Link
US (2) US6118822A (en)
EP (1) EP1033036B1 (en)
JP (1) JP2001525639A (en)
AT (1) AT232672T (en)
AU (1) AU1605299A (en)
DE (1) DE69811394T2 (en)
WO (1) WO1999029115A1 (en)

Families Citing this family (43)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4576975A (en) * 1983-03-03 1986-03-18 Minnesota Mining And Manufacturing Company Water soluble Michler's ketone analogs in combined photoinitiator composition and polymerizable composition
KR100311473B1 (en) * 1999-01-20 2001-11-02 구자홍 Method of search of optimal path for trellis based adaptive quantizer
US6593936B1 (en) * 1999-02-01 2003-07-15 At&T Corp. Synthetic audiovisual description scheme, method and system for MPEG-7
AU7859000A (en) 1999-10-05 2001-05-10 L-3 Communications Corporation A method for improving the performance of micromachined devices
US6738423B1 (en) * 2000-01-21 2004-05-18 Nokia Mobile Phones Ltd. Method for encoding and decoding video information, a motion compensated video encoder and a corresponding decoder
KR100856398B1 (en) * 2000-05-17 2008-09-04 삼성전자주식회사 Variable-length coding and decoding method and apparatus using plural mapping tables
JP2002204449A (en) * 2000-12-28 2002-07-19 Pioneer Electronic Corp Quantization unit setting device, quantization unit setting method, coder and coding method, and information recording medium
US6677868B2 (en) 2001-03-16 2004-01-13 Sharp Laboratories Of America, Inc. Entropy coding with adaptive syntax to replace high probability symbols with lower probabilities symbols
US20020168010A1 (en) * 2001-05-11 2002-11-14 Koninklijke Philips Electronics N. V. System and method for efficient automatic design and tuning of video processing systems
US7664182B2 (en) * 2001-08-31 2010-02-16 Panasonic Corporation Picture coding and decoding apparatuses and methods performing variable length coding and decoding on a slice header stream and arithmetic coding and decoding on a slice data stream
US7123656B1 (en) * 2001-10-01 2006-10-17 Realnetworks, Inc. Systems and methods for video compression
WO2003044963A1 (en) * 2001-11-16 2003-05-30 California Institute Of Technology Data compression method and system
DE10156851C1 (en) * 2001-11-20 2003-07-03 Infineon Technologies Ag Method and device for image compression
CN1275222C (en) * 2001-12-25 2006-09-13 株式会社Ntt都科摩 Signal encoding apparatus, signal encoding method, and program
US6763068B2 (en) * 2001-12-28 2004-07-13 Nokia Corporation Method and apparatus for selecting macroblock quantization parameters in a video encoder
US7206448B2 (en) * 2002-02-28 2007-04-17 At&T Corp. System and method for using pattern vectors for video and image coding and decoding
US7099387B2 (en) * 2002-03-22 2006-08-29 Realnetorks, Inc. Context-adaptive VLC video transform coefficients encoding/decoding methods and apparatuses
JP4368575B2 (en) 2002-04-19 2009-11-18 パナソニック株式会社 Variable length decoding method, variable length decoding apparatus and program
GB2393060B (en) * 2002-09-12 2006-01-25 Snell & Wilcox Ltd Manipulation of video compression
US7212681B1 (en) 2003-01-15 2007-05-01 Cisco Technology, Inc. Extension of two-dimensional variable length coding for image compression
EP1892965A3 (en) * 2003-04-04 2011-04-06 Avid Technology, Inc. Fixed bit rate, intraframe compression and decompression of video
US7403561B2 (en) * 2003-04-04 2008-07-22 Avid Technology, Inc. Fixed bit rate, intraframe compression and decompression of video
US7433519B2 (en) * 2003-04-04 2008-10-07 Avid Technology, Inc. Bitstream format for compressed image data
EP1629675B1 (en) * 2003-04-04 2008-01-09 Avid Technology, Inc. Fixed bit rate, intraframe compression and decompression of video
US7194137B2 (en) * 2003-05-16 2007-03-20 Cisco Technology, Inc. Variable length coding method and apparatus for video compression
US7454073B2 (en) 2004-06-15 2008-11-18 Cisco Technology, Inc. Video compression using multiple variable length coding processes for multiple classes of transform coefficient blocks
US7680349B2 (en) * 2004-08-18 2010-03-16 Cisco Technology, Inc. Variable length coding for clustered transform coefficients in video compression
US7454076B2 (en) * 2004-06-15 2008-11-18 Cisco Technology, Inc. Hybrid variable length coding method for low bit rate video coding
US7471841B2 (en) 2004-06-15 2008-12-30 Cisco Technology, Inc. Adaptive breakpoint for hybrid variable length coding
US20060062478A1 (en) * 2004-08-16 2006-03-23 Grandeye, Ltd., Region-sensitive compression of digital video
US7499596B2 (en) 2004-08-18 2009-03-03 Cisco Technology, Inc. Amplitude coding for clustered transform coefficients
US7471840B2 (en) * 2004-08-18 2008-12-30 Cisco Technology, Inc. Two-dimensional variable length coding of runs of zero and non-zero transform coefficients for image compression
US7499595B2 (en) * 2004-08-18 2009-03-03 Cisco Technology, Inc. Joint amplitude and position coding for photographic image and video coding
US7620258B2 (en) * 2004-08-18 2009-11-17 Cisco Technology, Inc. Extended amplitude coding for clustered transform coefficients
US7492956B2 (en) * 2004-08-18 2009-02-17 Cisco Technology, Inc. Video coding using multi-dimensional amplitude coding and 2-D non-zero/zero cluster position coding
JP4856954B2 (en) * 2005-06-08 2012-01-18 パナソニック株式会社 Image encoding device
JP4621574B2 (en) * 2005-10-06 2011-01-26 株式会社東芝 Image coding apparatus and image coding method
US7242328B1 (en) 2006-02-03 2007-07-10 Cisco Technology, Inc. Variable length coding for sparse coefficients
US8189676B2 (en) * 2007-04-05 2012-05-29 Hong Kong University Of Science & Technology Advance macro-block entropy coding for advanced video standards
US8902972B2 (en) * 2008-04-11 2014-12-02 Qualcomm Incorporated Rate-distortion quantization for context-adaptive variable length coding (CAVLC)
US8559742B2 (en) * 2008-10-10 2013-10-15 Accusoft Corporation Image encoding methods and apparatus providing improved visual results
ES2736003T3 (en) * 2011-04-22 2019-12-23 Dolby Int Ab Method and device for compression coding with data loss
GB2492392B (en) * 2011-06-30 2014-09-03 Canon Kk Method for encoding and decoding an image, and corresponding devices

Family Cites Families (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4716453A (en) * 1985-06-20 1987-12-29 At&T Bell Laboratories Digital video transmission system
US4945549A (en) * 1986-11-13 1990-07-31 The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration Trellis coded modulation for transmission over fading mobile satellite channel
US5297170A (en) * 1990-08-21 1994-03-22 Codex Corporation Lattice and trellis-coded quantization
US5233630A (en) * 1991-05-03 1993-08-03 Qualcomm Incorporated Method and apparatus for resolving phase ambiguities in trellis coded modulated data
US5231484A (en) * 1991-11-08 1993-07-27 International Business Machines Corporation Motion video compression system with adaptive bit allocation and quantization
KR0141824B1 (en) * 1991-12-23 1998-07-15 구자홍 Image compressing method & apparatus of variable length
US5329313A (en) * 1992-04-01 1994-07-12 Intel Corporation Method and apparatus for real time compression and decompression of a digital motion video signal using a fixed Huffman table
KR0150955B1 (en) * 1992-05-27 1998-10-15 강진구 Compressive and extensive method of an image for bit-fixing and the apparatus thereof
JP2940304B2 (en) * 1992-06-15 1999-08-25 松下電器産業株式会社 High efficiency coding apparatus, high efficiency coding method, and decoding apparatus for high efficiency coding apparatus
EP0574746A1 (en) * 1992-06-19 1993-12-22 Intel Corporation Method and apparatus for real time compression and decompression of a digital motion video signal using a fixed Huffman table
DE69605523T2 (en) * 1995-04-25 2000-07-06 Koninkl Philips Electronics Nv Device and method for encoding video images.
EP0768008B1 (en) * 1995-04-25 2000-07-05 Philips Electronics N.V. Device and method for coding video pictures
KR100207638B1 (en) * 1995-05-29 1999-07-15 윤종용 Variable bit rate video encoding method
US5724453A (en) * 1995-07-10 1998-03-03 Wisconsin Alumni Research Foundation Image compression system and method having optimized quantization tables
US5778192A (en) * 1995-10-26 1998-07-07 Motorola, Inc. Method and device for optimal bit allocation between different sources of information in digital video compression
TW364107B (en) * 1996-03-25 1999-07-11 Trw Inc Method and system for three-dimensional compression of digital video signals
US5745504A (en) * 1996-06-25 1998-04-28 Telefonaktiebolaget Lm Ericsson Bit error resilient variable length code
US5821887A (en) * 1996-11-12 1998-10-13 Intel Corporation Method and apparatus for decoding variable length codes
US6272175B1 (en) * 1997-02-13 2001-08-07 Conexant Systems, Inc. Video signal coding systems and processes using adaptive quantization
US6043844A (en) * 1997-02-18 2000-03-28 Conexant Systems, Inc. Perceptually motivated trellis based rate control method and apparatus for low bit rate video coding
US5974181A (en) * 1997-03-20 1999-10-26 Motorola, Inc. Data compression system, method, and apparatus

Also Published As

Publication number Publication date
DE69811394T2 (en) 2003-11-20
US6118822A (en) 2000-09-12
EP1033036A1 (en) 2000-09-06
JP2001525639A (en) 2001-12-11
US6249546B1 (en) 2001-06-19
AU1605299A (en) 1999-06-16
DE69811394D1 (en) 2003-03-20
AT232672T (en) 2003-02-15
WO1999029115A1 (en) 1999-06-10

Similar Documents

Publication Publication Date Title
US9774852B2 (en) Skip macroblock coding
US10455210B2 (en) Method, medium, and apparatus encoding and/or decoding an image using the same coding mode across components
US9843822B2 (en) Parameterization for fading compensation
US10045034B2 (en) System and method for using pattern vectors for video and image coding and decoding
US10659793B2 (en) DC coefficient signaling at small quantization step sizes
US8265405B2 (en) Image encoding apparatus, image decoding apparatus, image encoding method, image decoding method, image encoding program, and image decoding program
US6272180B1 (en) Compression and decompression of reference frames in a video decoder
US6763068B2 (en) Method and apparatus for selecting macroblock quantization parameters in a video encoder
CA2467496C (en) Global motion compensation for video pictures
DE69434271T2 (en) Adaptive variable length coding method for video data
US6134269A (en) Fixed or adaptive deinterleaved transform coding for image coding and intra coding of video
KR100953677B1 (en) Coding transform coefficients in image/video encoders and/or decoders
EP1853069B1 (en) Decoding macroblock type and coded block pattern information
US5428396A (en) Variable length coding/decoding method for motion vectors
Aravind et al. Image and video coding standards
EP0542474B1 (en) Adaptive coding and decoding of frames and fields of video signals
US6925119B2 (en) Process and system for the compression of digital video signals, a system and a computer program product therefor
US6590936B1 (en) Coded data transform method, transcoding method, transcoding system, and data storage media
EP1158809B1 (en) Variable length coding
KR100969645B1 (en) Adaptive variable length coding
KR100253931B1 (en) Approximate mpeg decoder with compressed reference frames
US5646618A (en) Decoding one or more variable-length encoded signals using a single table lookup
US5844613A (en) Global motion estimator for motion video signal encoding
EP1113673B1 (en) Quantization matrix for still and moving picture decoding
US6658157B1 (en) Method and apparatus for converting image information

Legal Events

Date Code Title Description
AK Designated contracting states

Kind code of ref document: A1

Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LI LU MC NL PT SE

17P Request for examination filed

Effective date: 20000630

17Q First examination report despatched

Effective date: 20001005

RAP1 Rights of an application transferred

Owner name: CONEXANT SYSTEMS, INC.

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: IT

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT;WARNING: LAPSES OF ITALIAN PATENTS WITH EFFECTIVE DATE BEFORE 2007 MAY HAVE OCCURRED AT ANY TIME BEFORE 2007. THE CORRECT EFFECTIVE DATE MAY BE DIFFERENT FROM THE ONE RECORDED.

Effective date: 20030212

Ref country code: GR

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20030212

Ref country code: NL

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20030212

Ref country code: FI

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20030212

AK Designated contracting states

Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LI LU MC NL PT SE

REG Reference to a national code

Ref country code: GB

Ref legal event code: FG4D

REG Reference to a national code

Ref country code: CH

Ref legal event code: EP

REG Reference to a national code

Ref country code: CH

Ref legal event code: NV

Representative=s name: ISLER & PEDRAZZINI AG

REF Corresponds to:

Ref document number: 69811394

Country of ref document: DE

Date of ref document: 20030320

Kind code of ref document: P

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: SE

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20030512

Ref country code: PT

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20030512

Ref country code: DK

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20030512

NLV1 Nl: lapsed or annulled due to failure to fulfill the requirements of art. 29p and 29m of the patents act
REG Reference to a national code

Ref country code: IE

Ref legal event code: FG4D

ET Fr: translation filed
PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: ES

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20030828

PGFP Annual fee paid to national office [announced from national office to epo]

Ref country code: SE

Payment date: 20031121

Year of fee payment: 6

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: CY

Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT

Effective date: 20031125

26N No opposition filed

Effective date: 20031113

PGFP Annual fee paid to national office [announced from national office to epo]

Ref country code: AT

Payment date: 20051102

Year of fee payment: 8

PGFP Annual fee paid to national office [announced from national office to epo]

Ref country code: MC

Payment date: 20051103

Year of fee payment: 8

PGFP Annual fee paid to national office [announced from national office to epo]

Ref country code: CH

Payment date: 20051125

Year of fee payment: 8

PGFP Annual fee paid to national office [announced from national office to epo]

Ref country code: IE

Payment date: 20051129

Year of fee payment: 8

PGFP Annual fee paid to national office [announced from national office to epo]

Ref country code: BE

Payment date: 20051208

Year of fee payment: 8

PGFP Annual fee paid to national office [announced from national office to epo]

Ref country code: LU

Payment date: 20051229

Year of fee payment: 8

PGFP Annual fee paid to national office [announced from national office to epo]

Ref country code: FR

Payment date: 20061117

Year of fee payment: 9

PGFP Annual fee paid to national office [announced from national office to epo]

Ref country code: GB

Payment date: 20061122

Year of fee payment: 9

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: AT

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20061125

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: IE

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20061127

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: LI

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20061130

Ref country code: CH

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20061130

Ref country code: MC

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20061130

Ref country code: BE

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20061130

PGFP Annual fee paid to national office [announced from national office to epo]

Ref country code: DE

Payment date: 20070102

Year of fee payment: 9

REG Reference to a national code

Ref country code: CH

Ref legal event code: PL

REG Reference to a national code

Ref country code: IE

Ref legal event code: MM4A

BERE Be: lapsed

Owner name: *CONEXANT SYSTEMS INC.

Effective date: 20061130

GBPC Gb: european patent ceased through non-payment of renewal fee

Effective date: 20071125

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: LU

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20061125

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: DE

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20080603

REG Reference to a national code

Ref country code: FR

Ref legal event code: ST

Effective date: 20080930

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: GB

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20071125

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: FR

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20071130